Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Aug 29;13(1):14135.
doi: 10.1038/s41598-023-41322-y.

Attention U-net for automated pulmonary fissure integrity analysis in lung computed tomography images

Zachary W Althof  1 , Sarah E Gerard  2 , Ali Eskandari  2 , Mauricio S Galizia  3 , Eric A Hoffman  1   2 , Joseph M Reinhardt  4   5
Affiliations

Attention U-net for automated pulmonary fissure integrity analysis in lung computed tomography images

Zachary W Althof et al. Sci Rep. .

Abstract

Computed Tomography (CT) imaging is routinely used for imaging of the lungs. Deep learning can effectively automate complex and laborious tasks in medical imaging. In this work, a deep learning technique is utilized to assess lobar fissure completeness (also known as fissure integrity) from pulmonary CT images. The human lungs are divided into five separate lobes, divided by the lobar fissures. Fissure integrity assessment is important to endobronchial valve treatment screening. Fissure integrity is known to be a biomarker of collateral ventilation between lobes impacting the efficacy of valves designed to block airflow to diseased lung regions. Fissure integrity is also likely to impact lobar sliding which has recently been shown to affect lung biomechanics. Further widescale study of fissure integrity's impact on disease susceptibility and progression requires rapid, reproducible, and noninvasive fissure integrity assessment. In this paper we describe IntegrityNet, an attention U-Net based automatic fissure integrity analysis tool. IntegrityNet is able to predict fissure integrity with an accuracy of 95.8%, 96.1%, and 89.8% for left oblique, right oblique, and right horizontal fissures, compared to manual analysis on a dataset of 82 subjects. We also show that our method is robust to COPD severity and reproducible across subject scans acquired at different time points.

PubMed Disclaimer

Conflict of interest statement

Drs. Hoffman and Reinhardt are shareholders in VIDA Diagnostics, Inc. Zachary Althof was formerly employed by VIDA Diagnostics, Inc. Drs. Gerard and Reinhardt have US patent 11,475,562 on the FissureNet fissure segmentation method. Drs. Galizia and Eskandari do not have any competing interests.

Figures

Figure 1
Figure 1
Fissures of the lungs. The right horizontal fissure separates the right upper lobe (RUL) and middle lobe (RML). The right oblique fissure divides the middle lobe (RML) and the lower lobe (RLL) anteriorly, and the RUL and RLL posteriorly. The left oblique fissure separates the left upper lobe (LUL) and lower lobe (LLL).
Figure 2
Figure 2
IntegrityNet pipeline. A CT image is input to the LungNet–FissureNet–LobeNet pipeline to obtain segmentations or the lungs, fissures, and lobes. During preprocessing the CT intensity values are clipped to (−1024, 200) and then linearly rescaled to (−1, 1). Next, the lung mask is used to compute a bounding box for each of the left and right lungs. The bounding boxes are used to crop the CT image and the fissure probability image to the left and right lungs. Lastly, the cropped CT and fissure probability images are concatenated along the channel dimension and the result is used as input to IntegrityNet. The output of each IntegrityNet is a fissure integrity mask where green and red represents intact and incomplete fissure, respectively.
Figure 3
Figure 3
IntegrityNet architecture. The left side of the network represents the contracting path where inputs are progressively down-sampled and features are extracted at each layer. The right side of the network represtents the expanding path where feature maps are progressively up-sampled to generate the final fissure integrity labels. The skip connections bring infromation from the contracting path into the expanding path to improve performance at each layer of up-sampling.
Figure 4
Figure 4
Left oblique fissure predicted fissure integrity score compared to ground truth. Red dashed line represents linear trendline.
Figure 5
Figure 5
Right oblique fissure predicted fissure integrity score compared to ground truth. Red dashed line represents linear trendline.
Figure 6
Figure 6
Right horizontal fissure predicted fissure integrity score compared to ground truth. Red dashed line represents linear trendline.
Figure 7
Figure 7
Example of a motion induced blurring artifact in a CT image.
See this image and copyright information in PMC

References

    1. Eberhardt R, Daniela G, Herth FJF, Schuhmann M. Endoscopic bronchial valve treatment: Patient selection and special considerations. Int. J. Chron. Obstruct. Pulmon. Dis. 2015;10(2147–57):8. doi: 10.2147/COPD.S63473. - DOI - PMC - PubMed
    1. Koster TD, Slebos DJ. The fissure: Interlobar collateral ventilation and implications for endoscopic therapy in emphysema. Int. J. Chron. Obstruct. Pulmon. Dis. 2016;11:765–773. doi: 10.2147/COPD.S103807. - DOI - PMC - PubMed
    1. Koster TD, et al. Predicting lung volume reduction after endobronchial valve therapy is maximized using a combination of diagnostic tools. Respiration. 2016;92:150–157. doi: 10.1159/000448849. - DOI - PubMed
    1. Sciurba FC, et al. A randomized study of endobronchial valves for advanced emphysema. N. Engl. J. Med. 2010;363(13):1233–1244. doi: 10.1056/NEJMoa0900928. - DOI - PubMed
    1. Galloy AE, Amelon RE, Reinhardt JM, Raghavan ML. Contact mechanics model of lung lobar sliding. Appl. Eng. Sci. 2022;10:100098.

Publication types